基于嵌入式机器视觉技术的血型自动识别系统  

Automatic Identification System of Blood Group Based on Embedded Machine Vision Technique

作　　者：廖强[1,2] 陈里里[1,2] 胡雪[3] 张国珍[3] 

机构地区：[1]重庆大学机械工程学院,重庆400030 [2]机械传动国家重点实验室,重庆400030 [3]重庆医科大学附属第一医院输血科,400016

出　　处：《现代科学仪器》2011年第6期51-55,63,共6页Modern Scientific Instruments

基　　金：重庆市卫生局科研项目(2010-2-086);重庆市自然科学基金(CSTC2006BB3176);重庆大学"211工程"三期创新人才培养计划建设项目(S-09106)

摘　　要：目的:为了实现自动检测血型,建立了基于嵌入式机器视觉的血型自动识别系统,并通过实验验证系统可行性和可靠性。方法:首先,依据血型检测的原理,以日本CCS同轴照明LED视觉检测光源、光源控制器、毛玻璃片、白色背景构成照明模块,以中星微301芯片的USB摄像头和mini2440开发板为核心,构建了集图像采集和分析于一体的嵌入式硬件模块。基于面向对象的思想,利用EVC(Embedded Visual C++)开发工具设计了系统的软件。然后,针对采集的微柱凝胶卡图像特点,通过试验分析,设计了具体的图像预处理算法。其中,运用小波变换完成图像对比度增强,运用遗传算法完成图像分割。在此基础上,设计了红细胞凝集物的特征参数提取算法。最后,利用BP神经网络构建了识别模型,完成对红细胞凝集物的类别判断,从而根据判别规则得到血型的检测结果。结果:多次试验结果表明:系统对-、+、++、+++、++++五类红细胞凝集物的判别正确率均为100%;能100%地正确检测出正常标本的ABO和Rh血型。结论:系统基本能满足临床对于血型检测自动化识别、准确性好、抗干扰能力强、稳定可靠等要求。Objective： In order to realize automatic detectionation for blood group,an automatic identification system（ATIS） based on embedded machine vision technique is established and the feasibility and reliability of the system are verified by its application.Method： First,based on the detection principle of blood group,the illuminating module was formed by CCS coaxial light sources,light source controller,ground glass and white background.By employing USB camera（Vimicro,301 microchip） and mini2440 development board as the core components,the embedded hardware module was designed to realize the image collection and processing.Then founded on object-oriented ideas,the system software was developed by EVC（Embedded Visual C＋＋） as a development tool.According to the image feature of the micro-column gel cassette assay,an new algorithm of imaging preprocessing was proposed.The wavelet transform was performed for image contrast enhancement,and genetic algorithm was adopted to realize the image segmentation.On this basis,we designed a new extraction algorithm for characteristic parameters of red blood cell agglutinator.Finally,the degree of agglutination in red blood cell suspension was determined automatically through BP neural network recognition model,and discriminant rule was used to determine blood group.Result： Experimental results indicated that the correct recognition ratios were 100% for the five degrees（-,＋,＋＋,＋＋＋,＋＋＋＋） of agglutination,and in the case of blood group determination,a correct recognition rate of 100% is achieved.Conclusion： This system is in better agreement with clinical requirements due to its high sensitivity,automatic recognition,high detection rate,as well as strong anti-jamming and stabilization.

关 键 词：嵌入式 机器视觉 血型 自动识别 BP神经网络 

分 类 号：R318.6[医药卫生—生物医学工程] TP391.41[医药卫生—基础医学]